Every 2050 risk number depends on which climate models you believe. Arasense scores each model against decades of observed weather at your location, then builds projections only from the models that earn their place, transparently, so a technical reviewer can audit every step.
For Bologna's worst rain day by mid-century, credible CMIP6 models span a huge range. Pick models blindly, or average everything, and your risk number is an accident.
Bologna's most extreme rain day gets lighter by 2050.
The same rain day gets over 40% heavier. Both models sit in the same CMIP6 ensemble.
Arasense answers it with measurement: each model's skill against 20 years of observed local climate.
Every CMIP6 model is compared against observed climate behaviour at your exact location and decomposed with the Aras Diagram: bias, variability, and phase alignment, not a single opaque score. Models that fail screening are rejected, and you see why.
Skill-weighted models estimate mid-century change for flood-driving rainfall, heat, and drought, with uncertainty ranges and model agreement reported, never hidden.
Each report names the trusted models, the rejected models, the weights, the spread, and the interpretation, in language a technical reviewer, auditor, or council can interrogate.
24 European cities ranked with one consistent method, plus a new global preview tier: world cities run through the same pipeline, each pin carrying a reference-data confidence flag. We trust-score our own baseline, not just the models.
Full ensemble screening, per-model weights, uncertainty range, and interpretation.
Sample evidenceSSP2-4.5 vs SSP5-8.5, multi-hazard profiles, portfolio ranking, validation notes.
Regional briefThe workflow applied to Puglia for adaptation, infrastructure, and advisory work.
A peer-reviewed framework decomposing model error into bias, variability, and phase alignment, supporting clearer judgement than any single aggregate score.
Diagnostics and projections built on established reanalysis and the full CMIP6 ensemble, with each model's skill made explicit.
Flood-screening pilots are compared against satellite-derived evidence windows, documenting where the workflow performs and where it remains uncertain.
The model-trust method behind Arasense is peer-reviewed and open. Aras’ Diagram Studio runs it live in your browser: paste observations and model series, and watch each model’s bias, variability, and correlation error resolve into one auditable picture.
The full KGE decomposition (correlation, bias ratio, variability ratio, and total error) computed instantly from your own CSV, with model ranking and publication-ready export. No installation, and your data never leaves the browser.
Izzaddin et al. (2024), Stochastic Environmental Research and Risk Assessment 38, 2261–2281, with an MIT-licensed reference implementation on GitHub.
Resilience planning, regional screening, and adaptation decisions backed by evidence that survives technical review, not dashboard scores nobody can explain.
Client deliverables and due-diligence work strengthened with transparent model-trust reasoning and clearly communicated uncertainty.
Pilot engagements begin with a geography, an asset class, or a single decision question, and produce a defensible evidence pack.
Arasense